线性混合效应模型预测值关于白噪声偏离的敏感性分析

本文应用Banerjee和Magnus于1999年提出的局部敏感性分析方法讨论了线性混合效应模型中预测值关于误差项白噪声偏离的敏感性问题,提出了敏感性度量统计量,并在AR(1)和MA(1)误差项条件下数值模拟了这些统计量的表现.结果表明,预测值在AR(1)和MA(1)误差项下一般都不具有明显的敏感性;特别是在MA(1)下预测值对误差项的白噪声偏离通常具有很好的稳健性.

On the Sensitivity of LME Predictors to Covariance Misspecification

This paper considers the sensitivity of the predictors in linear mixed-effects models to co- variance misspecification using the local sensitivity analysis proposed by Banerjee and Magnus.The sensitivity statistics are derived.Simulation studies show that the predictors are often insensitive to covariance misspecification under,especially disturbances which provides a useful reference to the prac- titioners.